|SUCCESS| [v5] examples/l3fwd: fix RFC 1812 checks

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: SUCCESS
https://dpdk.org/patch/172038

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-10-24

## DPDK Patch Review

### Errors

None.

### Warnings

None.

### Info

#### 1. Checksum validation logic improvement (correctness)

The revised checksum validation correctly handles all four `RTE_MBUF_F_RX_IP_CKSUM_*` states.

Line quoted:
```c
if (rte_ipv4_cksum(pkt) != 0)
```

The simplified checksum check (calculate without modifying the packet) is correct:
`rte_ipv4_cksum()` returns 0 when the checksum is valid.
This eliminates the prior bug where the checksum field was zeroed and not restored.

#### 2. Link layer length calculation (correctness)

The patch correctly subtracts the Ethernet header length when passing the link layer length to `is_valid_ipv4_pkt()`.

Lines quoted:
```c
if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt_len - sizeof(struct rte_ether_hdr),
		m->ol_flags) < 0) {
```

This fixes the prior bug where the full packet length (including L2) was compared against IP header fields that only account for L3+.

#### 3. IP header length field usage (correctness)

The patch correctly uses `rte_ipv4_hdr_len(pkt)` instead of `sizeof(struct rte_ipv4_hdr)` to account for IP options.

Line quoted:
```c
if (rte_be_to_cpu_16(pkt->total_length) < rte_ipv4_hdr_len(pkt))
```

This fixes the prior bug where IP options would cause valid packets to pass an invalid check.

#### 4. Additional RFC 1812 check (enhancement)

The new check comparing link layer length against IP total length follows RFC 1812 section 5.2.2 recommendations.

Lines quoted:
```c
if (link_len < rte_be_to_cpu_16(pkt->total_length))
	return -6;
```

This is a correct additional validation.

#### 5. Byte order conversion correctness

The patch correctly changes `rte_cpu_to_be_16()` to `rte_be_to_cpu_16()` when reading the `total_length` field.

Line quoted:
```c
if (rte_be_to_cpu_16(pkt->total_length) < rte_ipv4_hdr_len(pkt))
```

The `total_length` field is stored in network byte order (big-endian) in the IP header,
so converting from big-endian to CPU byte order is correct.
The prior use of `rte_cpu_to_be_16()` would only work correctly on big-endian hosts.

---

All identified bugs are fixed correctly.
The code follows DPDK style guidelines.
The logic is sound and improves upon the previous implementation.

Review-Result: CLEAN